A photo of a Philco from a CACM ad. Bob Lewis has (had?) a large framed picture but it is extremely faded and yellow. If scanned it would need quite a bit of image processing to be presentable.
This is reminiscent of IBM's T.J. Watson, who said "I think there's a world market for about five computers." Watson's son sharply disagreed, and was assisted in a way when the old man died. Thus IBM went forward and became a giant in the field.
The Philco was largely a gift, but came with a catch. The memory was an older style and could not be interleaved; part of the deal was that UW would develop a new box, called a PT10, to accomplish this using newer technology. This was done, including construction of new "breadboarding" equipment.
The card reader that came with the Philco proved to be fast but finicky. The "pick" finger wore quickly, and it was found that a good replacement could be made by carving up an old tire with a pocket knife. Even so, sometimes it would read 2000 CPM and sometimes it would destroy 2000 CPM. A donation allowed the purchase of a much better Uptime card reader, which was then re-worked to interface it with the Philco. This reader worked very well for the rest of its life, which encompassed both the Philco and the Sigma 7. It was rated at 1500 CPM but was operated at 1000 CPM for reliability.
Readers unfamiliar with early computing should be reminded that there were simply no disk drives. Period. No hard drives (except in IBM's laboratories and/or wealthy computing centers), no floppy drives. No terminals. Users stored their programs and data on punched cards, 7.5 x 3 inches (which matched the size of US money at the time Herman Hollerith invented the punched card for the US census in 1890). One punched card could hold 80 characters (one per column) or, in binary mode, 960 bits. Large electro-mechanical devices called card readers would feed the cards one at a time and read the cards (at first using metal brushes and later using light and photo-detectors). A programmer would sit at a keypunch and punch a program, or corrections (large corporations hired professional keypunchers to do this from sheets prepared by the programmers), and submit the deck to the operator. The operating system was booted from tape, and if the deck needed the Fortran compiler, this was loaded from tape as well, and so on. You really needed to be there to understand early computing.
The Philco had included a drum memory device which also was unreliable and was never seriously used. All processing was done on magnetic tape. An IBM-compatible 7-track tape drive was acquired and added to the system, allowing interchange with other systems around the country which had standardized on 1/2-inch tapes. Later this would be used to transfer data back and forth with the Sigma 7 and IBM 1401. The 1-inch tape drives had a capstan and used pinch-rollers for movement and braking, and would click/clack at each start/stop. Standing in the newer machine room in Biological Sciences, surrounded on both sides by the tape drives during a registration class-card sort, was an interesting experience; similar to being in a field of crickets. The tapes were also "formatted" like disks are, and records were always recorded in fixed locations. This had the advantage that a bad spot on a tape could be bypassed by formatting around it. It also had the "advantage" that you could re-write a block in the middle of a tape without re-writing or damaging the rest of the data. This proved to be a problem for one user who tried to do the same thing on the Sigma's standard 1/2-inch tapes, and found out that writing on a tape logically destroyed all following data (of course it doesn't impact *all* the data, but the erase head will erase tape for an inch or so beyond the new data, with unpredictable results thereon).
Bob Lewis tells me the main problem with the drum unit was that it would tend to forget what was stored on it. They would load the operating system and compilers, run fine for a day, then turn off the system for the night. The next morning, even after a considerable warm-up period, the data needed to be rebuilt again. Bob also describes the unit as truly massive, similar to the FASTRAND:
In the event of a power failure, the FASTRAND can be coupled to a standby generator for several days. Note that the total angular momentum available is slightly reduced if the data-storage option is fitted, owing to the braking effect of the read-write heads on the magnetized drum surface. Three or more FASTRANDs should not be switched on simultaneously at the same site without consulting Sperry's in-house geophysicist. The latter will also advise on the correct lattitude-dependent orientation of the drum axis to avoid data loss due to coriolis forces.
(reproduced from Stan Kelly-Bootle's book, The Devil's DP Dictionary, 1981 McGraw-Hill, under what I believe would be fair-use laws).
To help support Philco programming, we used to have an octal Monroe calculator. These were large mechanical desk-top machines that could add and subtract, and often multiply and divide. This unit was built to run in octal and lacked keys for the digits 8 and 9. I remember seeing it in the basement of Ivinson as recently as about 1982 or so, but it has since disappeared.
One of the more interesting quirks of the Philco was its time/date clock which was a mechanical gear train driven by a clock motor, coupled to a set of switches that the CPU could read. It kept track of the time, day of month, month, and I assume the year. It mechanically knew how many days each month had, except for the switch labeled LEAP YEAR. I believe the customer had to enable this switch as early as March 1 of a year prior to a leap year, or as late as February 28 of a leap year, and to then reset it as early as March 1 of the leap year or February 28 of the next year.
One of the frustrating aspects of working with the Philco was the need to use a keypunch. As time went by, use grew to the point where it was impossible to get time on a keypunch and there were signup sheets for them, too.